Neural-Network State-Of-Charge and State of Health Estimation

1. A method comprising: receiving, by a trained model, a time series of values of battery attributes for a battery that are determined based on measurements performed by one or more sensors monitoring the battery, wherein the battery attributes include: an ambient temperature, and one or both of a battery voltage and a battery current, and wherein a portion of the time series of values of battery attributes represents battery attributes of the battery during a charging cycle and another portion of the time series of values of battery attributes of the battery represents battery attributes of the battery during a discharging cycle; and generating, by the trained model, a battery state for the battery based on the time series of values, the battery state including a state of health (SOH) of the battery, wherein the trained model is a single, processor-based trained model that is trained to generate battery states for the battery during charging cycles of the battery and during discharging cycles of the battery.

2. The method of claim 1, wherein the trained model is a convolutional neural network (CNN).

3. The method of claim 1, wherein the battery attributes further include a state-of-charge (SOC) of the battery and wherein receiving, by the trained model, the time series of values of battery attributes includes receiving a time series of values of the SOC of the battery.

4. The method of claim 3, further comprising: determining, using a recurrent neural network (RNN), the time series of values of the SOC based on the battery voltage and the battery current; and providing the time series of values of the SOC to the trained model as one of the time series of values of battery attributes received by the trained model.

5. The method of claim 1, wherein the battery state generated by the trained model based on the time series of values further includes a state-of-charge (SOC) of the battery.

6. The method of claim 1, wherein the SOH represents a capacity of the battery.

7. The method of claim 1, further comprising: providing the battery state for controlling operation of the battery or an electrical apparatus coupled to the battery.

8. The method of claim 1, wherein the battery comprises a rechargeable battery, and wherein receiving the time series of values for the one or more battery attributes includes receiving a time series of values spanning multiple charging and discharging cycles of the rechargeable battery.

9. The method of claim 1, wherein generating the battery state includes determining the battery state at a first time dependent on the time series of values of the battery attributes spanning greater than 30 minutes of operation of the battery.

10. The method of claim 1, further comprising: setting values of parameters of the trained model based on processing a plurality of time series of values of the battery attributes and corresponding battery states, the plurality of time series of values spanning a plurality of different state-of-charge profiles.

11. The method of claim 1, wherein the trained model is a single neural network.

12. The method of claim 1, wherein the generating, by the trained model, of battery state for the battery comprises: generating, by the trained model when the battery is discharging, based on the time series of values, the battery state for the battery when the battery is discharging, and wherein the method further comprises: receiving, by the trained model, a further time series of values of the battery attributes for the battery; and generating, by the trained model when the battery is charging, based on the received further time series of values, a further battery state for the battery when the battery is charging.

13. A battery monitoring system comprising: one or more sensors configured to monitor a battery; and a processor coupled to the one or more sensors, the processor configured to: receive the sensor data from the one or more sensors; receive, at a trained model implemented by the processor, a time series of values of battery attributes for the battery that are determined based on measurements performed by the one or more sensors, wherein the battery attributes include: an ambient temperature, and  one or both of a battery voltage and a battery current, and wherein a portion of the time series of values of battery attributes represents battery attributes of the battery during a charging cycle and another portion of the time series of values of battery attributes of the battery represents battery attributes of the battery during a discharging cycle; and generate, via the trained model, a battery state for the battery based on the time series of values, the battery state including a state of health (SOH) of the battery, wherein the trained model is a single, processor-based trained model that is trained to generate battery states for the battery during charging cycles of the battery and during discharging cycles of the battery.

14. The system of claim 13, wherein the trained model is a convolutional neural network (CNN) comprising a plurality of layers, including at least one pooling layer and at least one fully connected layer.

15. The system of claim 13, wherein the time series of values of the battery attributes received by the trained model further include a time series of values of the state of charge (SOC) of the battery.

16. The system of claim 15, wherein the processor is further configured to: determine, using a recurrent neural network (RNN) implemented by the processor, the time series of values of the SOC based on the battery voltage and the battery current; and provide the time series of values of the SOC to the trained model as one of the time series of values of battery attributes received by the trained model.

17. The system of claim 13, wherein the battery state generated by the trained model based on the time series of values further includes a state-of-charge (SOC) of the battery.

18. The system of claim 13, wherein the battery comprises a rechargeable battery, and wherein the time series of values for the one or more battery attributes includes values spanning multiple charging and discharging cycles of the rechargeable battery.

19. A non-transitory computer readable media programmed with instructions, executable on one or more processors, to: receive, by a processor-based trained model, a time series of values of battery attributes for a battery that are determined based on measurements performed by one or more sensors monitoring the battery, wherein the battery attributes include: an ambient temperature, and one or both of a battery voltage and a battery current, and wherein a portion of the time series of values of battery attributes represents battery attributes of the battery during a charging cycle and another portion of the time series of values of battery attributes of the battery represents battery attributes of the battery during a discharging cycle; and generate, by the trained model, a battery state for the battery based on the time series of values, the battery state including a state of health (SOH) of the battery, wherein the trained model is a single, processor-based trained model that is trained to generate battery states for the battery during charging cycles of the battery and during discharging cycles of the battery.

20. The non-transitory computer readable media of claim 19, wherein the trained model is a convolutional neural network (CNN) comprising a plurality of layers, including at least one pooling layer and at least one fully connected layer.

21. The non-transitory computer readable media of claim 19, wherein the time series of values of the battery attributes received by the trained model further include a time series of values of the state of charge (SOC) of the battery.

22. The non-transitory computer readable media of claim 21, wherein the instructions, executable on the one or more processors, are further to: determine, using a recurrent neural network (RNN), the time series of values of the SOC based on the battery voltage and the battery current; and provide the time series of values of the SOC to the trained model as one of the time series of values of battery attributes received by the trained model.